USB interfaces and clocks

[vinylnvalves]

Trying to understand the benefits of changing the Amanero USB to I2S card in my DSP preamp, for something better. I heard a system recently with a reclocker in the path and it seemed to improve things. A bit confused as I thought a I2S interface into a DAC is asynchronous, so don’t understand why improving synchronicity is important?

If there is good reason to upgrade the I2S input, which approach technology is best? From the likes of the http://jlsounds.com/i2soverusb.html which an external clock can be added to, to the pink faun card at the other extreme.

For any internal card I have a 5v or 3.3v regulated supply available.

[Nick]

I used a Amanero card, FWIW. Not sure what reclocker would do with USB as the data transfer is in blocks. And if the Amanero is running async the receiver is specifying the receive rate anyway. I found using isolators between the Amanero and the DAC and between the i2s from the receiver and the DAC to make a worthwhile improvement. And also provided a simple way of switching inputs.

https://uk.rs-online.com/web/p/digital- ... cs/0459673

What is it that the something better does that is better than the Amanero?

[vinylnvalves]

I have galvanic isolation on the Amanero card already. Yes the question is whether someone else is better.. The JL sounds card has an external oscillator card that designed to go with it, so why both improving the clock if there is no benefit ?

The best card (allegedly) the Sonore audio bytes card https://www.sonore.us/legacy.html Made a thing about its better clocks.

[Nick]

As an aside, looking at the board you linked to. Does XMOS make chips for anything other than USB inputs boards? When I looked at doing my DAC i got a XMOS dev board, and one of the code sample was for a USB audio interface. It was just there as a use case, they expected the processor family to be used for a whole loads of real time and embedded systems and was intended to be a successor to the transputer. The USB sample seems to have been a lucky move as it may have been the only thing they get used for.

I never got my head around their language which was a mix of C, Occam, and some weird special use cases that made the chip IO work in many magic ways that was getting close to CPLD sort of flexibility. Checking their website they seem to be moving towards RISC-V, but still have the xcore processor and seem to be still going.

[Nick]

I have galvanic isolation on the Amanero card already

Ok, maybe a different card from what I used then, unless your isolation is off board. This is the board I used. https://amanero.com/

[Ray P]

I have galvanic isolation on the Amanero card already.

The difference between this arrangement, assuming I have it correct, and the JLSounds is that the JLSounds oscillators are on the DAC side of the galvanic isolation and are used to reclock the data right before the DAC. With your Amanero it would seem you are passing the clock and data signals through the isolator chips and my understanding is that they're poor with respect to jitter.

FWIW, I've used Amanero and JLSounds boards and the JLSounds is my go to for USB - it just sounds better. The JLSounds requires a good 5V supply to the 'clean side', with the 'dirty' side either powered via USB or, perhaps better, another 5V supply.

I typically use a Beaglebone Black based renderer though, but always with galvanic isolation and reclocking on the 'clean' side.

BTW, the standard JLSounds board already uses very good NKD onboard oscillators and there is some debate whether the add-on Crystek-equipped oscillator board actually delivers much/any real improvement - I've no experience to relate though. That said, their are lots of possibilities out there for 'hi-end' clocking solutions such as exotic oscillators, temperature compensation solutions etc. so having the ability to connect external oscillators would be useful but then you also get into details like the appropriate connectors to optimise your investment in that clocking...

[vinylnvalves]

I used the Twisted Pear Hermes card, for galvanic isolation.

[Nick]

With your Amanero it would seem you are passing the clock and data signals through the isolator chips and my understanding is that they're poor with respect to jitter.

Maybe, I was more interested in keeping ground noise out of the system. But you can do you. 1ns looks good enough for a clocked data stream but I worry more about the trailing wires that folk have in their DAC's than I probably should.

[Ray P]

I used the Twisted Pear Hermes card, for galvanic isolation.

That would be redundant with a JLSsounds board, or you could get a Twisted Pear Cronus reclocker to go between the Hermes and the DAC.

[vinylnvalves]

Thanks, for the info. I will have to digest the options. A bit distracted today, as had a fight with a ford ranger tow bar, and lost big style, so have to sort out my wheels before thinking about HiFi. That aside was thinking of ordering the JLsounds board and the oscillator board without the clocks, then can add the external clock later if necessary, most of the cost is in the crystek clocks, which you can get from Farnells anyway.

[vinylnvalves]

Car is in the insurance companies hands now, so can refocus on audio.

The other card people rave about is https://luckit.biz/. But I am not sure as the galvanic ground isn’t in the right place.

The system I heard that kicked off this was using a mutec to reclock and feed into spdif. In my mind the spdif is limited and technically shouldn’t be as good as the I2S input. Also wasn’t clear if the mutec was up sampling either - never been convinced upsampling does anything. But then I didn’t that how we dealt with digital input would make a difference.

Nick - I keep the leads short to the DAC input (<70mm). Was thinking next time I would use some shielded cable from an old usb lead.

Also what’s the thoughts about what contacts need to be connected - does the usb lead need to have the voltage wire if the I2S card is externally powered? As I would have thought if it’s not needed it’s better without, I know a lot of the boutique usb cables put the voltage wire external to the signal cables.

[Nick]

I keep the leads short to the DAC input (<70mm)

That's about 0.25ns for reference. To my mind that's not a trivial distance for high speed signals.

Was thinking next time I would use some shielded cable from an old usb lead.

Then you will need to find the characteristic impedance of the cable and make sure its terminated correctly.

[vinylnvalves]

70mm is probably shorter than the DAC boards input pins to the chip. I take it from your comment that shielded cable isn’t necessary?

[Nick]

70mm is probably shorter than the DAC boards input pins to the chip

OK, so that may or may not be a good thing, depends on the board, but for reference 70mm is bigger than the entire digital board in the DAC I made including the mount for the Amanero board. You can get a lot of stuff in that space. Have a look at the effort digital folk make to ensure signals arrive at the same time. But TBH, the sort of data rate that audio requires is such that its all a lot lot simpler than a processor interface to RAM. Doesn't stop it being used as a selling point by marketing though. Clocked data line i2s just needs the data to be there when the clock changes state. 13MHz for 192k i2s has a clock period of something like 80ns, so as long as the data and clock are within 20ns of each other it should be fine.

I take it from your comment that shielded cable isn’t necessary

It entirely depends on what you want to do and the distance you want to do it over. Ethernet went from 10Mbs over a coax cable, to 10Gbs over twisted pair, so there is no coax good, wire bad thing.

[vinylnvalves]

I suppose I can put the I2S board ( which ever one I go for) on the 2.5mm pins of the DAC/DSP board. I didnt as the pin call outs weren’t common . But can work out some crossover pin arrangement. Will end up having longer wires to the I2S board, but an extra 200mm in a 1m usb cable, when you use upto 2.5m is probably no issue.